Release 0.3.0

CHANGELOG.md

@@ -15,6 +15,14 @@ a section is written.
 
 ## [Unreleased]
 
+Nothing yet.
+
+## [0.3.0] - 2026-09-23
+
+The configuration applies as you save it, a broken one says so, the games
+you marked yourself in the Game Bar are followed too, and the watcher costs
+a twelfth of what it did at rest.
+
 ### Added
 
 - The watcher reads `config.toml` again whenever you save it, within a
@@ -29,11 +37,11 @@ a section is written.
   *Edit configuration* opens the file, and saving a good one brings the
   icon back, with a notification saying the watcher is watching again —
   also when the fix came while the watcher was stopped.
-
 - Games you marked yourself with *Remember this is a game* in the Game Bar
   are detected like any other, from their launch — or from the moment you
-  tick the box, in the middle of a game. Windows never tells the rest of the
-  system about those, which is why they used to go unnoticed.
+  tick the box, in the middle of a game. Windows does not start, for those,
+  the signal the program used to watch alone, which is why they went
+  unnoticed.
 - At start, the log says which of the games you marked Microsoft's own list
   now knows, so you can untick them and let Windows recognise them itself.
   `status` lists the games you marked and says the same.
@@ -130,6 +138,7 @@ and when it stops, on Windows' own signal that a game is running.
   including the scheduled tasks a program that needs administrator rights
   requires, and a Windows power plan.
 
-[Unreleased]: https://github.com/Geeooff/GameModeExecutor/compare/v0.2.0...HEAD
+[Unreleased]: https://github.com/Geeooff/GameModeExecutor/compare/v0.3.0...HEAD
+[0.3.0]: https://github.com/Geeooff/GameModeExecutor/compare/v0.2.0...v0.3.0
 [0.2.0]: https://github.com/Geeooff/GameModeExecutor/compare/v0.1.0...v0.2.0
 [0.1.0]: https://github.com/Geeooff/GameModeExecutor/releases/tag/v0.1.0

Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "game-mode-executor"
-version = "0.2.0"
+version = "0.3.0"
 authors = ["Geoffrey Vancoetsem"]
 edition = "2024"
 # Let chains, stable since 1.88.

docs/design/08-distribution.md

@@ -304,6 +304,29 @@ only proposed:
   shortcut is a package change and would go with the recipes artefact if
   that is taken; the reload is the next lot's.
 
+## The package that held one piece — 2026-09-23
+
+Preparing 0.3.0, the local `build.ps1 release` said *release OK*: 102 ICE
+evaluators clean, the changelog section found, the binaries stamped. The
+installer weighed 88 KB, against 1.5 MB for 0.2.0. Its `files.cab` stream
+was 20,378 bytes, for two executables of 1.4 MB each.
+
+`makecab`'s default disk is a 1.44 MB floppy. The two executables had grown
+by some 50 KB each since 0.2.0 and compressed, together, to 1,476,966
+bytes — just over one floppy, where 0.2.0's had fitted by twenty
+kilobytes. So `makecab` spanned two cabinets, and with a fixed
+`CabinetNameTemplate` the second overwrote the first under the same name.
+The package kept the last piece; the `File` table still listed all four
+files, which is all the evaluators check, and `msi.ps1` only checked that
+a cabinet existed. Published, it would have failed on every machine — and
+the first to meet it would have been the updater, on the update meant to
+close [Lot 13](13-updating.md).
+
+`msi.ps1` now asks for one cabinet whatever its size (`MaxDiskSize=0`),
+and refuses a package whose cabinet does not list, through `expand -D`,
+every file the package installs, or whose build left more than one
+cabinet. Caught by looking at a number that had no reason to change.
+
 ## Versioning
 
 Decided 2026-09-17, when the question came up before publication: why the

scripts/msi.ps1

@@ -95,10 +95,16 @@ $work = Join-Path ([System.IO.Path]::GetTempPath()) "gamemode-executor-msi-$([gu
 New-Item -ItemType Directory $work | Out-Null
 try {
     # --- the cabinet ---------------------------------------------------------
+    # MaxDiskSize=0: one cabinet, however large. makecab's default disk is a
+    # 1.44 MB floppy, and with a fixed CabinetNameTemplate the second cabinet
+    # of a spanned set overwrites the first under the same name: on
+    # 2026-09-23 the 0.3.0 executables compressed to 1,476,966 bytes, just
+    # over, and the package came out at 88 KB holding the last 20 KB piece.
     $ddf = @(
         '.OPTION EXPLICIT',
         '.Set CabinetNameTemplate=files.cab',
         ".Set DiskDirectoryTemplate=$work",
+        '.Set MaxDiskSize=0',
         '.Set CompressionType=LZX',
         '.Set Cabinet=on',
         '.Set Compress=on',
@@ -116,6 +122,13 @@ try {
     $null = & makecab.exe /F $ddfPath
     $cab = Join-Path $work 'files.cab'
     if (-not (Test-Path $cab)) { throw 'makecab produced no cabinet' }
+    # The cabinet must hold every file, whole: its existence proved nothing
+    # the day it held one piece of two. `expand -D` lists what it contains.
+    $listed = @(& expand.exe -D $cab | ForEach-Object { if ($_ -match ':\s+(\S+)\s*$') { $Matches[1] } })
+    foreach ($f in $files) {
+        if ($listed -notcontains $f.Key) { throw "the cabinet lacks $($f.Key) ($($f.Path)); makecab listed: $($listed -join ', ')" }
+    }
+    if (@(Get-ChildItem $work -Filter '*.cab').Count -ne 1) { throw 'makecab produced more than one cabinet' }
 
     # --- the database --------------------------------------------------------
     $installer = New-Object -ComObject WindowsInstaller.Installer